Contrasting invertebrate immune defense behaviors caused by a ...

RESEARCH ARTICLE Open Access

Contrasting invertebrate immune defensebehaviors caused by a single gene, theCaenorhabditis elegans neuropeptidereceptor gene npr-1Rania Nakad1,2, L. Basten Snoek3, Wentao Yang1, Sunna Ellendt1, Franziska Schneider1, Timm G. Mohr1,Lone Rösingh1, Anna C. Masche1, Philip C. Rosenstiel4, Katja Dierking1, Jan E. Kammenga3

and Hinrich Schulenburg1*

Abstract

Background: The invertebrate immune system comprises physiological mechanisms, physical barriers and alsobehavioral responses. It is generally related to the vertebrate innate immune system and widely believed to providenonspecific defense against pathogens, whereby the response to different pathogen types is usually mediated bydistinct signalling cascades. Recent work suggests that invertebrate immune defense can be more specific at least atthe phenotypic level. The underlying genetic mechanisms are as yet poorly understood.

Results: We demonstrate in the model invertebrate Caenorhabditis elegans that a single gene, a homolog of themammalian neuropeptide Y receptor gene, npr-1, mediates contrasting defense phenotypes towards two distinctpathogens, the Gram-positive Bacillus thuringiensis and the Gram-negative Pseudomonas aeruginosa. Our findingsare based on combining quantitative trait loci (QTLs) analysis with functional genetic analysis and RNAseq-basedtranscriptomics. The QTL analysis focused on behavioral immune defense against B. thuringiensis, using recombinantinbred lines (RILs) and introgression lines (ILs). It revealed several defense QTLs, including one on chromosome Xcomprising the npr-1 gene. The wildtype N2 allele for the latter QTL was associated with reduced defense against B.thuringiensis and thus produced an opposite phenotype to that previously reported for the N2 npr-1 allele against P.aeruginosa. Analysis of npr-1 mutants confirmed these contrasting immune phenotypes for both avoidance behaviorand nematode survival. Subsequent transcriptional profiling of C. elegans wildtype and npr-1 mutant suggested thatnpr-1 mediates defense against both pathogens through p38 MAPK signaling, insulin-like signaling, and C-type lectins.Importantly, increased defense towards P. aeruginosa seems to be additionally influenced through the induction ofoxidative stress genes and activation of GATA transcription factors, while the repression of oxidative stress genescombined with activation of Ebox transcription factors appears to enhance susceptibility to B. thuringiensis.

Conclusions: Our findings highlight the role of a single gene, npr-1, in fine-tuning nematode immune defense,showing the ability of the invertebrate immune system to produce highly specialized and potentially opposing immuneresponses via single regulatory genes.

Keywords: Caenorhabditis elegans, Pathogen avoidance behavior, Innate immunity, Immune specificity, QTL analysis

* Correspondence: [email protected] of Evolutionary Ecology and Genetics, Zoological Institute,University of Kiel, 24098 Kiel, GermanyFull list of author information is available at the end of the article

© 2016 Nakad et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Nakad et al. BMC Genomics (2016) 17:280 DOI 10.1186/s12864-016-2603-8

BackgroundIn contrast to higher vertebrates, which have adaptive im-mune response systems, invertebrates exclusively rely onthe innate immune system (the immune system is here de-fined sensu lato as the organism’s defense against infection,including avoidance behavior, physical barriers, and physio-logical processes). For a long time it was assumed that onlythe adaptive system is capable of mounting highly specificdefense responses. However, evidence is accumulating thatinvertebrates have surprisingly complex immune systemsthat in theory may have the potential to produce similarspecificities [1–3]. Yet, to date, we possess only little in-formation on the genetic and molecular mechanismsunderlying such specificities. First insights into thesemechanisms were previously obtained for the modelnematode C. elegans, an important invertebrate systemfor studying immune defense [2, 4, 5]. For example, lossof function of the prolylhydroxylase encoding gene egl-9 enhances susceptibility to Staphylococcus aureus [6]but resistance to Pseudomonas aeruginosa (PA01) [7]and Bacillus thuringiensis toxins [8]. Similarly, a loss offunction of the Toll-like receptor gene tol-1 increasessusceptibility to Salmonella enterica but resistance toEnterococcus faecalis [9], even though the general im-portance of tol-1 in worm immunity is unclear [5, 10].Such specificities may not only be expressed by the

nematode’s physiological immune system, but could alsobe expected for behavioral defenses. Such behaviors are acentral component of immune defense sensu lato - nextto protective barriers and physiological processes - andare likely to represent a highly economic immune defensestrategy because they simultaneously reduce pathogencontact, and thus the risk of tissue damage, and also thenecessity to activate the energetically costly physiologicaland cellular response [11]. C. elegans colonizes microbe-rich habitats in nature where it feeds on bacteria andyeasts [12–15]. Since these habitats also contain manypathogenic microorganisms, C. elegans has evolved dis-tinct types of behavioral responses including physicalavoidance, associative learning and reduced oral uptake ofpathogens [4, 16–22].Previous studies revealed the presence of substantial

genetic variation among wild isolates of C. elegans intheir behavioral response towards different pathogens[17, 19, 23–27]. In one case, namely the defense responseagainst the Gram-negative bacterium Pseudomonas aeru-ginosa, this variation could be linked to the polymorphicneuropeptide receptor npr-1 locus on the X chromosome.The gene npr-1 was proposed to regulate C. elegans’ im-munity against PA14 either through controlling the aero-taxis response [17], or through controlling both aerotaxisresponse and physiological immune defense [18]. npr-1 isa homolog of the mammalian neuropeptide Y receptorgene and it is found in two different isoforms in C. elegans

that result from a single amino acid change at position 215(valine in isoform 215 V; phenylalanine in isoform 215 F)[28]. These isoforms do not only influence pathogendefense but also foraging behavior in response to oxygenconcentrations [28, 29] and leaving behavior from lawnswith the laboratory food bacterium Escherichia coli [30, 31].The apparent complexity of the C. elegans defense

against pathogens [1–3, 5] raises the question whether sin-gle pathways or genes can also fine-tune the behavioraldefense response towards specific pathogens. To addressthis question we studied the genetic architecture of behav-ioral immune defense of C. elegans towards the Gram-positive pathogen Bacillus thuringiensis. This pathogen islikely to coexist with C. elegans in nature [15]. Some strainsare nematocidal, whereby the host is infected by the oraluptake of spore-toxin mixtures. Infection of the gut isfollowed by toxin-mediated cellular damage of the intes-tinal epidermis, germination of spores and subsequentproliferation of vegetative cells, including expression ofvarious virulence factors, ultimately resulting in nematodedeath [32–36]. Nematocidal B. thuringiensis induces pro-nounced behavioral responses in C. elegans [21, 23, 37, 38].Here we explored genetic variation in C. elegans and

used quantitative trait locus (QTL) analysis to characterizethe genetic basis of behavioral immune defense againsttwo pathogenic B. thuringiensis strains, whereby onestrains (BT B-18679) is known to be more pathogenic thanthe other (BT B-18247) [39, 40]. Our QTL analysis wasbased on a panel of 200 recombinant inbred lines (RILs)and 90 introgression lines (ILs), derived from a cross be-tween the C. elegans strains N2 and CB4856 [41, 42]. OurQTL analysis identified npr-1 as one of the candidategenes, though with an opposite effect on avoidance be-havior to that previously reported towards P. aerugi-nosa [17, 18]. Therefore, we further characterized thefunction of the npr-1 gene in producing contrasting patho-gen defense responses. Using npr-1 mutants, we assessedthe influence of the gene on both avoidance behavior andsurvival towards the two pathogen species, B. thuringiensisand P. aeruginosa. Moreover, we used RNAseq to identifydifferences in the pathogen-dependent transcription ofnpr-1 down-stream targets. The functional importanceof such differences was assessed through enrichmentanalysis of gene ontology (GO) categories, customizednematode-specific gene sets, which we collated from pre-vious gene expression analyses, and transcription factorbinding motifs.

Results and discussionTwo C. elegans wild-type strains differ in bacterial lawnleaving behaviorThe standard laboratory strain N2 and the Hawaiian strainCB4856 showed significant variation in lawn-leaving be-havior towards nematocidal B. thuringiensis strains (B-

Nakad et al. BMC Genomics (2016) 17:280 Page 2 of 20

18679 & B-18247) and non-nematocidal strains (DSM350& E. coli OP50). Lawn leaving served as a proxy for behav-ioral defense and was based on an assay (Additional file 1)related to those previously used to characterize C. elegansavoidance behavior [10, 16, 19, 21, 31, 37], in this caseusing peptone-free medium (PFM) to prevent B. thurin-giensis spore germination outside the host (see Methods).Lawn leaving behavior was significantly higher for CB4856compared to N2 on all tested bacterial strains and for allexposure time periods (Fig. 1; Sheet 1 in Additional files 2,3 and 4). For both C. elegans strains, we observed a signifi-cant increase in leaving across time (Fig. 1). For both, theavoidance response towards the most pathogenic strain(B-18679) was higher than that towards the less patho-genic strain (B-18247) (Fig. 1c, d).Our results confirm previously reported higher avoid-

ance behavior and resistance of CB4856 compared to N2towards one of the pathogens used in the current study,B. thuringiensis B-18247 [23]. Our findings are also con-sistent with two previous studies that demonstrated a

higher OP50-patch leaving behavior [31] and a highermicrosporidia resistance of CB4856 compared to N2[43]. Interestingly, the opposite phenotype has been re-ported regarding the nematode’s response to two otherpathogens, P. aeruginosa and Serratia marcescens. Inthese cases, N2 rather than CB4856 produced higher re-sistance and behavioral avoidance towards P. aeruginosa[17, 19], and higher avoidance towards S. marcescens[26]. Moreover, as the more pathogenic B-18679 wasmore strongly avoided than the less virulent B-18247(Fig. 1c, d), C. elegans appears to be able to differenti-ate between different levels of pathogenicity of thesame bacterial species. In this case, the difference inpathogenicity is likely due to expression of differentCry toxins that result in different infection patterns[37]. Based on our results we expect that the N2 andCB4856-derived RIL and IL populations are likely tocontain sufficiently high levels of variation for a QTLanalysis of avoidance behaviors towards the fourchosen bacterial strains.

Fig. 1 Variation in lawn leaving behavior between N2 and CB4856 towards different bacterial strains. Lawn leaving behavior towards: (a) thenon-pathogenic E. coli strain OP50; (b) the non-nematocidal B. thuringiensis strain DSM350; (c) the nematocidal B. thuringiensis strain B-18247;and (d) the highly nematocidal B. thuringiensis strain B-18679. The Y axis shows the leaving index for N2 (blue) or CB4856 (red), calculated asthe proportion of nematodes which left the bacterial lawn; the X axis indicates the different exposure time points. Asterisks denote the treatments forwhich leaving behavior of CB4856 is significantly different from N2. The dotted reference line shows a leaving index of 0.5. Error bars representstandard error of the mean. Raw data are given in sheet 1 of Additional file 2. Detailed statistical results are shown in Additional file 3

Nakad et al. BMC Genomics (2016) 17:280 Page 3 of 20

Multiple QTLs and their interactions account for variationin avoidance behaviorWe performed QTL analyses on C. elegans pathogendefense and revealed the genetic architecture of patho-gen avoidance behavior to (i) be polygenic, (ii) includeepistatically interacting loci, and (iii) incorporate generalas well as pathogen-specific avoidance loci. In particular,our study simultaneously assessed the behavioral re-sponse of 200 RILs and 90 ILs [41, 42] towards four bac-terial strains (two nematocidal B. thuringiensis and twonon-nematocidal controls) at two exposure time points(14 h and 24 h) and with three replicates per treatmentcombination, using the same lawn leaving assays asabove for N2 and CB4856. Below we present our resultsof a main-effect QTL analysis of the RIL population andan analysis of interaction effects among loci for the RILpopulation.The main effect QTL analysis uncovered five main re-

gions associated with avoidance: (i) one large region in themiddle of chromosome II, for which the CB4856 allele(s)increase(s) leaving behavior of the non-pathogenic bac-teria only at the 24 h time point (Fig. 2a, b); (ii) a regionon chromosome II, for which the N2 allele(s) specificallyincrease(s) avoidance of the more pathogenic strain B-18679 at the 24 h time point (Fig. 2d); (iii) a region on theleft arm of chromosome IV, for which the N2 allele(s) in-crease(s) leaving behavior towards controls and pathogens(Fig. 2a-d); (iv) a region on the right arm of chromosome

IV, for which the CB4856 allele(s) increase(s) avoidance ofpathogens at both time points (Fig. 2c, d); and (v) a regionon chromosome X with the strongest effect on leaving be-havior towards controls and pathogens (Fig. 2a-d), medi-ated by the CB4856 allele(s) and including a significanttime effect on the response to the pathogenic bacteria(Fig. 2c, d). This very strong X chromosome effect wasconfirmed by the ILs (Additional file 5).Our analysis of interaction effects among QTLs, using a

standard interaction model (phenotype ~ time +marker1 *marker2), revealed several significant intra-genomic asso-ciations with an influence on lawn leaving behavior. For E.coli, significant interactions were found for at least twocases (Fig. 3a): (i) between the beginning of chromosome Iand the first half of chromosome X; and (ii) between theend of chromosome II and almost the entire IV chromo-some. For DSM350, we identified interaction effects be-tween: (i) the end of chromosome IV and the first quarterof chromosome X; and (ii) the end of chromosome II andalmost the entire IV chromosome (Fig. 3b). The latterseems to be specific for food patch leaving behavior as itwas identified for both of the non-pathogenic bacteria(Fig. 3a, b). For the nematocidal B-18247, we found inter-action effects at least between: (i) the beginnings ofchromosome II and V; and (ii) the second quarter ofchromosome II and the beginning of chromosome X(Fig. 3c). For the highly nematocidal B-18679, severalinteraction effects were identified including: (i) between

Fig. 2 QTL profiles for single marker mapping of avoidance behavior. QTL analysis of avoidance of: (a) E. coli OP50; (b) the non-nematocidal B.thuringiensis DSM350; (c) nematocidal B-18247; and (d) highly nematocidal B-18679. The X-axis shows the markers along the five autosomes andthe X chromosome. Vertical light gray lines denote the boundaries between chromosomes. The Y-axis indicates the association between thechromosomal markers and variation in avoidance. Significance is indicated by –log10 of the p-value obtained from the linear model, which ismultiplied by the sign of the effect to indicate the N2 allelic effect on avoidance. A value above 0 indicates an increase in leaving caused by theN2 allele; a value below 0 indicates an increase in leaving caused by the CB4856 allele. By convention, values above +2 or below -2 are considered toindicate a significant influence. Green and red lines show the results after either 14 h or 24 h exposure, respectively

Nakad et al. BMC Genomics (2016) 17:280 Page 4 of 20

the ends of chromosome I and II; and (ii) between thesecond quarter and the middle of chromosome IV(Fig. 3d). Interestingly, the X chromosome region withthe strongest influence in the main effect model (Fig. 2)only contributed to very few significant interaction ef-fects. One of these is an interaction with a chromosomeI region,mediating avoidance of E. coli OP50 (Fig. 3a),and another one with a chromosome IV region, influ-encing avoidance of DSM350 (Fig. 3b).Taken together, our results demonstrate that pathogen

avoidance has a complex genetic architecture in C. ele-gans, which overlaps with, but differs from the response tonon-pathogenic microbes. In particular, pathogen defensetraits are related to the response to non-pathogenic bac-teria, because they are affected by the same loci. Defenseis thus in part determined by the general response to mi-crobes, whereby pathogenicity of the bacteria may simplyelevate the response mediated by a particular locus, as

indicated for the X chromosome QTL (Fig. 2). Moreover,our results for the pathogen-specific QTLs are consistentwith the previous finding that pathogen defense in inver-tebrate animals seems to rely on few loci and involve epi-static interactions among them [44, 45], possibly as aconsequence of reciprocal coevolution among host andpathogens [44]. It will be a rewarding challenge for the fu-ture to characterize the genes underlying the pathogen-specific QTLs. Interestingly, the main effect QTL on theX chromosome was previously implicated in lawn leavingbehavior with similar allelic effects towards the non-pathogenic E. coli OP50 (i.e., the CB4856 allele increasesavoidance; [31]) but with opposite allelic effects to-wards P. aeruginosa (i.e., the N2 allele increases avoid-ance; [17–19]). In these cases, the QTL effect onchromosome X could be associated with variation in thegene npr-1 [17–19, 31] and, at least towards E. coli, add-itionally the catecholamine receptor gene tyra-3 [31].

Fig. 3 Heat-map of interaction effects for avoidance behavior. Results for avoidance of: (a) E. coli OP50; (b) DSM350; (c) nematocidal B-18247; and(d) nematocidal B-18679. The distribution of markers across the genome is shown on both axes. The chromosome boundaries are indicated bythe thin vertical and horizontal lines. A color legend for significance of the interaction between markers is shown on top of the panels. Significance isin –log10(p). High significance values are given in “warm” colors with purple and red indicating the highest significances, followed by orange and thenyellow. For example, in panel (c) showing the results for B-18247, the thin red area towards the bottom right indicates a significant interaction betweenloci from the beginning of the X chromosome (horizonal axis) and the middle of chromosome II (horizontal axis). The other slightly larger red area inthe bottom right of this panel points to a significant interaction between the beginning of chromosome V (horizontal axis) and the beginningof chromosome II (vertical axis)

Nakad et al. BMC Genomics (2016) 17:280 Page 5 of 20

The npr-1 gene affects defense against B. thuringiensisThe gene npr-1 was previously linked to the strong-effectQTL on chromosome X [17], which we found in thecurrent study to influence C. elegans defense towards B.thuringiensis in the opposite way than that towards P. aer-uginosa. We therefore specifically tested whether this geneis indeed responsible for the contrasting phenotypes inpathogen avoidance and resistance.We first studied the role of npr-1 in lawn leaving behav-

ior on E. coli and B. thuringiensis using two mutants (npr-1(ad609), npr-1(ur89)), which were both previously shownto decrease pathogen avoidance behavior against P. aerugi-nosa [17, 18]. Another gene from the left arm of chromo-some X was previously demonstrated to influence foodpatch leaving behavior, namely tyra-3, which encodes a tyr-amine receptor homologue [31]. Therefore, we furthertested its involvement in pathogen defense with theknock-out mutant tyra-3(ok325).Analysis of the two mutant npr-1 alleles (npr-1(ad609),

npr-1(ur89)) yielded different results in avoidance and

resistance (Figs. 4 and 5; sheets 2–4 of Additional files 2, 6,7 and 8). In particular, avoidance behavior of npr-1(ur89)was similar to CB4856 but higher than that of N2 (Fig. 4,Additional file 6). In contrast, npr-1(ad609) always pro-duced a low leaving rate, similar to N2 and clearly differentfrom CB4856. On the highly pathogenic B-18679, avoid-ance behavior was extremely high at both time pointswithout any significant differences among the C. ele-gans strains (Fig. 4d). In addition, the tyra-3 mutantconsistently showed a similar behavioral response toN2, irrespective of the bacterium and the exposure time(Fig. 4, Additional file 6).The npr-1 alleles produced similarly contrasting effects

on survival rate, which is often used as a proxy for nema-tode immunity. For the RIL/IL parental strains, we foundthat CB4856 showed significantly higher resistance thanN2 on both nematocidal B. thuringiensis strains (Fig. 5;Additional files 7, 8 and 9). Moreover, the npr-1(ur89)mu-tant was significantly more resistant than N2 and as resist-ant as CB4856 on both nematocidal pathogens, whereas

Fig. 4 Lawn leaving behavior for N2, CB4856, and different mutants towards E. coli and B. thuringiensis. Results for avoidance of: (a) E. coli OP50;(b) non-nematocidal B. thuringiensis DSM350; (c) nematocidal B. thuringiensis B-18247; and (d) highly nematocidal B-18679. The Y axis shows theproportion of escaped worms (i.e., leaving index); the X axis shows the two time points at which leaving was scored. The asterisk (*) is shown forthe treatments which differ significantly from N2 whereas the hash sign (#) indicates significant differences from CB4856. The dotted referenceline shows a leaving proportion of 0.5 for orientation among sub-panels. Error bars represent standard error of the mean. Raw data are given insheet 2 of Additional file 2; detailed statistical results in Additional file 6

Nakad et al. BMC Genomics (2016) 17:280 Page 6 of 20

npr-1(ad609) was as susceptible as N2 on both pathogenicstrains (Fig. 5, Additional file 8). None of the C. elegansstrains showed any mortality under control conditions (re-sults not shown).We conclude that the two mutant npr-1 alleles produce

opposite effects on both behavioral avoidance of the fourbacterial strains and also resistance against nematocidal B.thuringiensis. Consequently, variation in npr-1 may onlypartially explain the strong main effect QTL on the Xchromosome. The difference between the two npr-1 allelesin avoidance and resistance of B. thuringiensis is surprising,because both alleles behaved similarly in previous studiesinvestigating resistance against P. aeruginosa [17, 18]. Yetthe two alleles carry different mutations: npr-1(ad609) twoin exons 2 and 3, whereas the mutation of npr-1(ur89) fallsinto exon 3 (http://www.wormbase.org). The exact reasonsfor the different effects of these alleles clearly deserve fur-ther investigation in the future, ideally including additionalloss-of-function and also reduced-function npr-1 alleles incombination with a tissue-specific analysis of the muta-tional effects. We further conclude that the tyra-3 genedoes not appear to influence the assayed phenotypes(Figs. 4 and 5), including avoidance of E. coli OP50, whichwas however previously demonstrated in a separate study[31]. The difference in results could be due to variation inexperimental approaches. For example, we directly charac-terized leaving behavior, whereas the previous study scoredactivity as a proxy for leaving behavior [31].

Contrasting effect of npr-1 on defense againstPseudomonas aeruginosaWe sought to confirm the previously published finding[17–19] that the wild-type N2 produces higher resistanceand stronger avoidance behavior towards the pathogen P.aeruginosa PA14 than the Hawaiian strain CB4856 and twonpr-1 mutants [17–19], thus contrasting with our above re-sults for B. thuringiensis. Here, we specifically re-evaluatedthese previous results under our laboratory conditions and

assay protocols, using the peptone-rich NGM plates re-quired for expression of P. aeruginosa virulence. We firstused the lawn leaving assay to assess the avoidance re-sponse against PA14 at different exposure time points.Consistent with previous findings, the npr-1 mutants andCB4856 showed significantly lower PA14 pathogen avoid-ance than N2 across all time points (Fig. 6; sheets 5 and 6of Additional files 2, 10, 11 and 12). For the 48 h time pointthe mutant npr-1(ad609) even had a lower leaving responsethan CB4856 (Fig. 6b). On OP50, leaving behavior wassimilar for all C. elegans strains at all time points except attime point 14 h, when the mutant npr-1(ad609) showed amore pronounced leaving behavior than N2 (Fig. 6a,Additional file 12). In this assay, we also included atyra-3 mutant, which expressed a similar leaving re-sponse to N2 under all conditions except at time point48 h, where its leaving response against PA14 was reducedin comparison to N2, but still significantly higher thanthat of the remaining strains (Fig. 6b, Additional file 12).We next evaluated the effect of npr-1 on resistance

against PA14 using standard C. elegans survival assays. Allstrains survived less on the pathogen PA14 than on thecontrol OP50 (Fig. 7; sheet 7 of Additional files 2 and 13).On the pathogen, N2 was significantly more resistant thanall other tested strains (Fig. 7b, Additional file 14).We conclude that npr-1 directly influences avoidance of

PA14, in agreement with previous work [17–19]. In thesestudies, npr-1 was suggested to affect PA14 resistance ei-ther as a consequence of hyperoxia avoidance behavioronly (proposed by Reddy et al., [17]) or through bothhyperoxia avoidance and the regulation of physiologicalimmune responses (proposed by Styer et al., [18]).Our results further demonstrate that the two C. elegans

wild-type strains express opposite phenotypes on the twotested pathogens and that this contrast may be mediatedat least partially by npr-1, as one of the npr-1 alleles alsoproduces an opposite phenotype relative to N2. Such op-posite effects in the wild-type strains indicate specific

Fig. 5 Survival of N2, CB4856, npr-1(ad609) and npr-1(ur89) on nematocidal B. thuringiensis. Results for: (a) nematocidal B-18247; and (b) highlynematocidal B-18679. Survival on the Y axis was plotted against relative B. thuringiensis concentration on the X axis. Error bars represent standarderror of the mean. The raw data are provided in sheet 4 of Additional file 2; the statistical results in Additional file 8

Nakad et al. BMC Genomics (2016) 17:280 Page 7 of 20

interactions with pathogens. The underlying genetics forsuch specificities have not yet been explored for behavioralimmune defense. Some information is available for physio-logical and cellular immune specificities. In the higher ver-tebrates, such specificities can be mediated by componentsof the adaptive immune system such as the highly variablereceptors of the major histocompatibility complex or thehighly variable T and B cell receptors. Similar specificitieshave also been recorded in invertebrates [1, 46], wherethey may be due to different immune signaling cascades.For example, in Drosophila, the immune deficiencypathway appears to be more important in the systemicresponse to Gram-negative bacteria, whereas the Tollpathway is more important towards Gram-positive bac-teria and fungi [47]. Moreover, in C. elegans, mutationsin the egl-9 and tol-1 gene enhance resistance againstsome pathogens, while simultaneously increasing suscepti-bility to other pathogens (see introduction and [5–8, 10]).Our study thus provides one of the few examples which

demonstrate that a single gene, in this case the neurope-tide Y receptor homolog gene npr-1, produces contrastingpathogen specificities in an invertebrate.At the same time, it is less clear how exactly npr-1

causes these contrasting phenotypes. Previous work onnematode social behavior demonstrated that npr-1 influ-ences worm aggregation, lawn bordering and clumpingthrough its effect on aerotaxis behavior. The two testednpr-1 alleles and also that of the CB4856 strain result ina preference towards lower oxygen concentrations usu-ally found at the edge of the bacterial lawn [48, 49],whereas the N2 allele shows no such preference. A simi-lar difference in aerotaxis behavior may also explain thereduced P. aeruginosa avoidance and resulting highersusceptibilities of the CB4856 and npr-1 mutant strains,which remain in longer contact with the harmful patho-gen, because the bacterial lawn boundaries show thepreferred lower oxygen concentrations [17]. An involve-ment of such aerotaxis behavior in the B. thuringiensis

Fig. 6 Lawn leaving behavior for N2, CB4856 and different mutants towards E. coli and P. aeruginosa. Results for: (a) E. coli OP50; and (b) P. aeruginosastrain PA14. The Y axis shows the proportion of escaped worms; the X axis shows the three time points at which leaving was scored (12 h, 24 h, and48 h). The asterisk (*) denotes the treatments which differ significantly from N2, whereas the hash sign (#) indicates significant differences from CB4856.The dotted reference line shows a leaving proportion of 0.5. Error bars represent standard error of the mean. The raw data are provided in sheet 6 ofAdditional file 2; detailed statistical results in Additional file 12

Fig. 7 Survival of N2, CB4856, npr-1(ad609) and npr-1(ur89) on E. coli and P. aeruginosa. Results for: (a) E. coli OP50; and (b) P. aeruginosa strainPA14. Survival (Y axis) was plotted against time in hours (X axis). We used the Kaplan-Meier method to calculate survival fractions. The raw dataare given in sheet 7 of Additional file 2; statistical results in Additional files 13 and 14

Nakad et al. BMC Genomics (2016) 17:280 Page 8 of 20

response of CB4856 and one of the npr-1 mutants wouldthen require that the preferred oxygen concentration isoutside of the bacterial lawn, which is unlikely to be thecase (assuming higher oxygen concentrations outside,where no oxygen is consumed by proliferating bacteria).Thus, it is conceivable that CB4856 and the one npr-1mutant are directly responding to a compound producedby B. thuringiensis, and that this response is less pro-nounced in the N2 strain and the other npr-1 mutant. Inthis context, it is worth noting that the high variationbetween N2 and CB4856 in their leaving response towardsthe control E. coli OP50 was only observed on peptonefree PFM but not the peptone-rich NGM assay plates(Figs. 1a and 4a versus Figs. 6a and Additional file 10).This is most likely explained by bacterial proliferation,which is possible on NGM but not PFM assay plates. Inturn, the lack of proliferation on the PFM plates is unlikelyto coincide with large variation in oxygen concentration,such that an aerotaxis response should be less pronouncedunder these conditions. Yet, a non-proliferating, staticbacterial population may produce particular metabo-lites, which then could have induced the CB4856 avoid-ance response.

npr-1 influences the transcriptomic response to B.thuringiensis and P. aeruginosaTo explore the mechanisms underlying the npr-1 medi-ated contrasting effects on immune defense, we assessedwhether npr-1 differentially affects gene expression in thepresence of either of the two pathogens. Using RNAseq wecompared the transcriptomes of the N2 and npr-1(ur89)strains exposed to either the nematocidal B. thuringiensisB-18247, the pathogenic P. aeruginosa PA14, or the controlE. coli OP50. We chose the mutant npr-1(ur89) because itshowed differential leaving behavior and survival on bothpathogens compared to N2 (Figs. 4 and 6). Exposure exper-iments were performed on Agar plates fully covered withbacterial lawns, thus reducing possible avoidance behaviorsand producing comparable levels of lawn occupancy forthe worms from the various treatment combinations. RNAtranscript levels were characterized at two time points,12 h and 24 h of pathogen exposure. We used principalcomponent analysis (PCA) to explore which experimentalfactors generated different transcriptional responses. Thefirst principal component indicated that the two nematodestrains vary in their transcriptional signature to all threebacteria (Fig. 8). The second principal component high-lights variation across several additional factors. The stron-gest effect stems from exposure time (light versus darkcolors; Fig. 8). Additional influences can be seen for patho-gen exposure versus the corresponding control, especiallyat the later time point (filled versus open symbols of thesame type; Fig. 8) and also a clearly distinct signal after24 h exposure to PA14 compared to all other conditions

(filled dark colored circles towards the bottom of thegraph; Fig. 8). These latter differences are more pro-nounced for N2 than the npr-1 mutant, especially asN2 produces clearly distinct treatment signatures at thelater 24 h time point (i.e., clearly separated dark blueopen and filled circles and squares; Fig. 8). One possiblereason for lower differentiation in the npr-1 mutantmay be a lower number of differentially expressed genescompared to the N2 strain. This was indeed the case, es-pecially upon pathogen exposure (Table 1), suggesting thatmutations in npr-1 somehow compromise the signallingresponse to pathogen infection.To identify groups of co-regulated genes, we next per-

formed K-means clustering on the significant gene sets.The resulting eight clusters confirm that the transcrip-tional response is influenced by the C. elegans strain, thepathogen strain and also the exposure time point (Fig. 9;Additional file 15). In detail, clusters 1, 2, 3, and 4 referto genes with strong differential expression upon expos-ure to only pathogenic B. thuringiensis B-18247 in onlythe C. elegans N2 strain, but neither the npr-1(ur89) mu-tant on the same pathogen nor any of the other treat-ments with P. aeruginosa (e.g., the stronger the colorintensity in Fig. 9, the stronger the expression differencebetween pathogen versus non-pathogen exposure). Thisresult again highlights that the npr-1 mutant shows gener-ally lower responsiveness in inducible gene expression(i.e., most clusters do not show high color intensity inFig. 9). Clusters 1, 2, 3 and 4 only responded to the patho-gen B-18247, and clusters 7 and 8 only or at least predom-inantly to PA14. Two clusters are specific to expressionvariation at the 12 h time point, in both cases upon

Fig. 8 Principal component analysis of transcriptomic variation.Variation is assessed for npr-1(ur89) and N2 upon exposure to pathogens(B-18247 and PA14) and the control bacterium OP50. N2 is shown inblue whereas npr-1(ur89) is shown in red. Light and dark colors indicatethe early and late time point, respectively (12 h versus 24 h). Filled andopen symbols denote exposure to pathogens (B-18247 and PA14) andcontrol bacteria (OP50), respectively. NGM is the nematode growthmedium enriched with peptone for the PA14 assay plates, PFMrefers to peptone-free NGM used for the B-18247 assay plates

Nakad et al. BMC Genomics (2016) 17:280 Page 9 of 20

exposure to B-18247 (i.e., clusters 1 and 3), whereas fourclusters indicate a more pronounced response at the later24 h time point, either towards only B-18247 (clusters 2and 4) or only PA14 (clusters 7 and 8). The remaining twoclusters highlight patterns of early or continuous tran-scriptional response towards B-18247 and late transcrip-tional response towards PA14 (clusters 5 and 6; Fig. 9).None of the identified clusters showed an opposite geneexpression pattern between either N2 and the npr-1 mu-tant (e.g. up in N2 and down in the npr-1 mutant) or thetwo pathogens (e.g. up after B. thuringiensis but down afterP. aeruginosa exposure). Taken together, clusters 5 and 6appear to encompass a general defense response againstboth pathogens, whereas the clusters 1, 2, 3, and 4 definethe specific response to B-18247 and clusters 7 and 8 thatto PA14. Therefore, the latter two groupings are likely toaccount for the observed npr-1 dependent defense differ-ences towards the two pathogens. We thus conclude thatthe considered mutation in the npr-1 gene causes a de-creased transcriptomic response to the two pathogens,which induce overlapping and distinct sets of differentiallyexpressed genes.

Different functions and signaling processes are affectedby the pathogen-dependent npr-1-specific transcriptomeWe used enrichment analysis as a statistical tool to explorethe possible functions of the differentially regulated geneclusters. Four types of enrichment analyses were performed,which aim to identify significant over-representation of(i) genes with a specific gene ontology (GO) term (GOterm analysis); (ii) customized nematode-specific genesets, inferred from previous gene expression analysesand based on the program EASE (EASE analysis); (iii) geneswith specific transcription factor-binding motifs (Motif ana-lysis), and (iv) expression QTLs (eQTLs). The customizedenrichment analysis with the program EASE [50] was basedon a large database of all previous C. elegans transcriptome

studies, WormExp [51], which we collated from publishedwork. These studies investigated differential gene expres-sion (i) across development, (ii) in specific tissues, (iii) inworms with defined mutations or subjected to RNAi-knockdown of specific genes, or (iv) upon exposure to en-vironmental stimuli such as pathogens, heavy metals, andother chemical compounds [51]. The GO term and Motifanalyses were based on published methods, such as DAVID[52, 53] and AMD [54]. Analysis of eQTL enrichment ex-pression differences, using the eQTL database collated fromdifferent previous eQTL analyses, all based on RIL panels

Table 1 Number of up- and down-regulated genes in the N2and npr-1(ur89) strains

B-18247 PA14

C. elegans 12 h 24 h 12 h 24 h

Up-regulated

N2 1393 1094 222 529

npr-1(ur89) 179 191 188 352

Down-regulated

N2 2384 458 81 1233

npr-1(ur89) 117 70 58 370

RNAseq was used to assess variation in gene expression among the C. elegansN2 and npr-1(ur89) mutant strain in response to exposure to nematocidal B.thuringiensis B-18247 and P. aeruginosa PA14, always relative to the respectiveE. coli OP50 control. Gene expression variation was studied at two time points,12 h and 24 h after initial exposure. The results are shown separately for theup- and down-regulated genes (top and bottom part of the table, respectively)

Fig. 9 Co-regulation of the differentially expressed genes. K-meanscluster analysis yielded eight clusters of co-regulated genes, as indicatedby the numbers on the left. The top of the graph shows the differenttreatment conditions, including C. elegans strain, pathogen strain andexposure time point. Red refers to up-regulated genes, whereas blue todown-regulated genes, always upon pathogen exposure relative to thecorresponding E. coli OP50 control. High color intensities indicate strongexpression differences to the control (see legend in the bottom rightcorner). The complete results are given in Additional file 15

Nakad et al. BMC Genomics (2016) 17:280 Page 10 of 20

derived from N2 and CB4856 as parental lines [41, 55–60],thus potentially allowing us to link the identified QTLs tothe expression variation inferred against pathogens. The re-sults are summarized in Figs. 10 and 11, Table 2, Additionalfiles 16 and 17, and explained in more detail below.We first focus on the general defense response against

the two pathogens, which is defined by two clusters (i.e,clusters 5 and 6; Fig. 9). These clusters are enriched forgenes previously implicated in C. elegans pathogen defense.These include genes involved in carbohydrate binding

(Fig. 10a), most likely mediated by C type lectin-like genes,many of which underlie this GO term (Additional file 17under GO) and which are up-regulated across treatmentconditions (Fig. 11a) and have repeatedly been implicatedin C. elegans immunity, possibly as pathogen recognitionreceptors or antimicrobials [11, 61–64]. These two clustersare also enriched for genes which were previously shownto respond to exposure to the same pathogens and othertypes of stressors, such as heavy metals, osmotic stress, orpesticides (Fig. 10b). The upregulated genes appear to be

Fig. 10 Functional consequences of gene expression variation between npr-1(ur89) and N2 upon pathogen exposure. a Enrichment of geneontology (GO) terms. The shown terms were significant with FDR < 0.05 (Additional file 17). b Overview of enrichment of pathogen- and stress-induced gene sets, inferred from EASE analysis on the various clusters of co-regulated genes. The significantly enriched gene sets are indicatedon the top and include - from left to right - differentially expressed genes upon exposure to (i) B. thuringiensis Cry5B toxin [89]; (ii)-(iii) the sameB-18247 strain used in the current study in C. elegans isolate MY15 or MY18 [90]; (iv)-(v) the same PA14 strain used here [66, 91]; (vi) Oxidativestress response [92]; (vii)-(ix) Osmotic induction [93]; (x)-(xi) Heavy metal Cadmium dysregulated genes [89, 94]; (xii) Pesticide influence [58]. c Overviewof enriched gene sets for selected immunity pathways and general categories, including (i)-(ii) the p38 MAPK pathway (pmk-1 and sek-1 targets; [91]);(iii)-(v) insulin signalling (daf-2 targets; [95–97]); (vi) npr-1 targets [18]; (vii) Glycoproteins; (viii)-(x) Cytochrome P450 [98–100]; (xi) Protein kinase [98]; (xii)Lipid metabolism [98]; (xiii) Cell division [98]. d Enrichment of Ebox and GATA motifs and their transcriptional targets. The enriched gene sets wereinferred with EASE and are indicated at the top, including differentially expressed genes in mutants of (i)-(iii) E-box transcriptions factors [101] or (iv)-(v)GATA transcription factors ELT-2 and ELT-3 [102, 103]. Enriched transcription factor binding motifs were inferred with AMD and are shown on the right(Additional file 17). In all panels, the clusters are given on the very left and are identical to those in Fig. 9. Red color indicates an enrichment for up-regulated genes per gene set, blue that for down-regulated genes per gene set. Color intensity corresponds to the significance level, inferred by EASEanalysis (see scale at the right side)

Nakad et al. BMC Genomics (2016) 17:280 Page 11 of 20

controlled by two of the main C. elegans immunity signal-ling cascades, the p38 MAPK and the insulin-like receptorpathways (Fig. 10c) [2], and also the npr-1 gene (Fig. 10c)[18]. They also show an enrichment in their promotor se-quences for a GATA binding motif, although not for knowntargets of the GATA transcription factors ELT-2 and ELT-3(Fig. 10d). They are also enriched for gene sets defined byeQTLs on chromosome I, III, the middle of chromosomeIV, and the left arm and the middle of chromosome X(Table 2). One of the enriched X chromosome eQTLs en-compasses the npr-1 gene, another the gene sek-1 of thep38 MAPK cascade, and the one on chromosome IV mayinclude the MAPK gene jnk-1 or the p38 homolog pmk-1,additionally supporting the role of these genes in the nema-tode’s expression response. The enriched eQTLs fromchromosome IV and the left arm of the X chromosome alsolie within the QTLs identified to influence behavioraldefense against B. thuringiensis (Fig. 2). Taken together,we conclude that clusters 5 and 6 comprise the compo-nents of a general defense response, apparently activenot only against pathogens but also other stressors andmediated by central stress and immune response path-ways. In the npr-1 mutant, this defense response isstrongly reduced towards both pathogens.The specific response to P. aeruginosa is captured by two

clusters (i.e., clusters 7 and 8; Fig. 9). They are enriched foreQTLs on chromosome I, V, and X (Table 2), although in

Fig. 11 Example for enriched Gene ontology terms. a Heatmap forcarbohydrate binding on cluster 5; b for oxidation reduction oncluster 7. Red and blue color indicate up- and down- regulationcompared to OP50

Table 2 eQTL enrichment analysis on identified expressionclusters

Expressioncluster

Location of eQTL eQTL set

Chromosome Approximate position

1 V 3.7 M Rockman

2 IV 6.6 M Viñuela (old)

3 X 15.5 M Rockman

4 - - -

5 X 2.3 M Rockman

X 5.8 M Rockman

X 10.9 M Rockman

IV 6.3 M Rockman

III 1.9 M Rockman

6 I 5.0 M Viñuela (juvenile)

7 X 15.7 M Rockman

V 12.3 M Rockman

8 I 3.9 M Rockman

eQTL enrichment analysis was performed to identify significant overlapsbetween the genes underlying a specific cluster in our analysis (first column;see also Fig. 9) and previously characterized gene sets that define particulareQTLs (last column). Such overlaps can then be linked to the specific QTLregions within the genome (second and third column), which may thencontain regulatory elements important for the expression variation in ourstudy and which may also have been identified as QTLs for the observedvariation in behavioral immune defense (Figs. 2 and 3). For further details seeAdditional file 16. No significant enrichment was found for cluster 4

Nakad et al. BMC Genomics (2016) 17:280 Page 12 of 20

chromosomal regions without any known regulator of im-mune defense. These clusters, especially the upregulatedcluster, are however enriched for genes previously shownto respond to infection by the same pathogen (Fig. 10b)and also those involved in the response to oxidativestress and xenobiotics, including cadmium and pesti-cides (Fig. 10a, b). The response to oxidative stress andxenobiotics is dominated by an up-regulation of cyto-chrome P450 genes (Fig. 11b; Additional file 17). Theup-regulated cluster is further influenced by the twomain immunity pathways, the npr-1 gene, and also aGATA transcription factor (Figs. 10c, d). All of the lattercomponents have previously been shown to be central forimmune defense against P. aeruginosa, especially the p38MAPK signalling cascade [65] and the GATA transcrip-tion factor ELT-2 [66]. The transcriptomic response to P.aeruginosa infection is additionally strongly influenced bycytochrome P450 expression, possibly as part of a generalstress response to reduce oxidative stress (Fig. 10c). Be-cause these responses are activated more strongly in theN2 strain, they are likely to mediate the observed higherresistance and behavioral defense for this strain comparedto the npr-1(ur89) mutant (Figs. 6 and 7 and model inFig. 12).The specific response to B. thuringiensis is defined by

four clusters, two up-regulated and two down-regulatedgroups of genes (i.e., clusters 1–4; Fig. 9). They are

enriched for eQTLs in the middle of chromosome IV(Table 2), which could contain known defense regulatorsagainst B. thuringiensis, the MAPK genes jnk-1 and pmk-1[67], and which lies within the QTL above identified tocontribute to behavioral defense against this pathogen(Fig. 2). Enriched eQTLs are additionally found on the leftarm of chromosome V and the right arm of the X chromo-some (Table 2), in both cases without a link to any knownimmune regulator. Moreover, the upregulated clustersshow an over-representation of genes involved in meta-bolic processes and phosphatase activity (Fig. 10a). Theyalso include genes known to respond to the same patho-gen and cadmium, as well as genes that are usually down-rather than up-regulated upon osmotic and pesticidestress (Fig. 10b). These two upregulated clusters areenriched for an Ebox transcription factor binding motifand the corresponding targets (Fig. 10d). The two down-regulated clusters show an enrichment for oxidationreduction and developmental genes (Fig. 10a), genes re-sponsive to B. thuringiensis and cadmium (Fig. 10b), andgenes controlled by insulin-like signalling, including glyco-proteins (Fig. 10c). One of the down-regulated clustersalso shows an enrichment of the GATA transcription fac-tor targets (Fig. 10d).Taken together, the results of our enrichment analyses

allow us to propose a possible mechanistic basis for thecontrasting defense effects of the npr-1 gene. It is worth

Fig. 12 Model of npr-1 dependent effects on pathogen defense in the N2 C. elegans strain. Exposure to both pathogens leads to an npr-1 dependentactivation of carbohydrate-binding factors, such as C-type lectin-like proteins, and also two central immune signaling cascades, the p38 MAPK and theinsulin-like pathways, which could all enhance pathogen resistance (middle part of the graph). Upon exposure to P. aeruginosa PA14 (right side), npr-1also influences the activation of a general stress response, via one or several GATA transcription factors(s), which increases oxidative stress resistanceand thus resistance to the pathogen. The response to B. thuringiensis (left side) is mediated by npr-1 through one or several Ebox transcription factors,resulting in a reduced oxidative stress response and increased metabolic activity as a possible cause of enhanced pathogen susceptibility. Arrows withlight colors indicate uncertain connections

Nakad et al. BMC Genomics (2016) 17:280 Page 13 of 20

reiterating that N2 and the mutant only differ in thepresence of a mutation in the npr-1 gene. Therefore, thedifferences observed between strains must be influencedby the allelic variation in this gene. The higher resistanceand avoidance behavior of N2 towards P. aeruginosa islikely influenced by the activation of GATA and/or p38MAPK targets and/or the induced oxidative stress re-sponse (see above; Fig. 12). The situation is less clear forthe response to B. thuringiensis. Because N2 produceslower defenses against B. thuringiensis than the npr-1(ur89) mutant (Figs. 4 and 5), and because any differen-tial gene expression is repressed in the npr-1(ur89) mutant(Table 1; Fig. 9), the specific activation of certain genes inN2 (i.e., for the two up-regulated clusters 1 and 2) and/orthe suppression of other gene groups in N2 (i.e., thedown-regulated clusters 3 and 4) must account for the ob-served lower resistance and avoidance response againstthis pathogen. We hypothesize that this may possibly bemediated by one of the following two processes or a com-bination thereof (see model in Fig. 12): (i) the lower oxida-tive stress response in N2 could lead to increasedsusceptibility towards B. thuringiensis, in analogy to theeffect recently described towards E. faecalis [68] and as-suming that B. thuringiensis toxicity causes oxidativestress (which is however currently unknown); and/or (ii)an activation of metabolic processes could be disadvan-tageous during pathogen infection, because it may re-duce availability of energetic resources that can beinvested in immune defense and because metabolicproducts may be exploited as a source of nutrition bythe pathogen. Any of the other implicated functions(Fig. 10) may also contribute to enhanced susceptibilityin an as yet unknown manner. These processes thenseem to be influenced by npr-1 through Ebox-specifictranscription factors. Interestingly, the higher suscepti-bility is apparently caused by an activation of specificfunctions and signalling processes rather than their ab-sence or at least reduced activity. This may indicate asub-optimal response to this specific pathogen in the N2strain or represent a consequence of pathogen-mediatedmanipulation of host responses, which are widespreadamong pathogens [69] and which have also been shownfor another Bacillus species, Bacillus nematocida, tochange C. elegans behavior and intestinal responses [70].At the moment, it is unclear in what way the indicatedprocesses influence either behavioral or physiological re-sponses or both simultaneously. This represents a challen-ging topic for future research.

ConclusionOur study revealed a complex genetic architecture com-prising several epistatically interacting QTLs associatedwith variation in C. elegans pathogen avoidance behavior.The most significant QTL encompassed the gene npr-1.

Our functional analyses of this gene revealed a contrastingeffect of npr-1 on C. elegans immune defense, as assessedthrough both behavioral and also survival phenotypes. Inparticular, the CB4856 allele was associated with fasterlawn leaving behavior and higher survival than the N2allele on B. thuringiensis, whereas it was associated withlower lawn leaving behavior and lower survival on P. aeru-ginosa. A further characterization of the exact role of npr-1 suggested that it mediates differential regulation ofdefense genes via either GATA transcription factors lead-ing to increased immune defense towards P. aeruginosa orEbox transcription factors leading to decreased immunedefense towards B. thuringiensis. Our study thus demon-strates that a single gene in C. elegans mediates contrast-ing pathogen-specific defense responses.

MethodsC. elegans and bacterial strainsC. elegans strains: (i) the standard wild-type strains N2and CB4856; (ii) 200 Recombinant Inbred Lines (RILs)and 90 Introgression lines (ILs) generated from crossesbetween N2 and CB4856 [41, 42, 55, 60]; and (iii) two dis-tinct mutant alleles of npr-1, npr-1(ur89) X (strain IM222)and npr-1(ad609) X (strain DA609), and also the tyra-3knock-out allele tyra-3(ok325) X (strain VC125). The threemutant strains were obtained from the CaenorhabditisGenetics Center (CGC; http://www.cbs.umn.edu/CGC/)and were all generated in the N2 background. All wormstrains were maintained at 20 °C on Nematode GrowthMedium (NGM) plates with the non-pathogenic E. coliOP50 as an ad libitum food according to standard proto-cols [71]. All mutants were backcrossed at least threetimes to N2. Presence of the target mutation was con-firmed for the two npr-1 mutants by sequencing the npr-1gene at the location of the mutations and for the knock-out mutant tyra-3 by polymerase chain reaction (PCR)analysis of the deleted region.Bacterial strains: (i) two nematocidal strains of B. thurin-

giensis, NRRL B-18247 and NRRL B-18679, originally pro-vided by the Agriculture Service Patent Culture Collection(United States Department of Agriculture, Peoria, Illinois,USA); (ii) the non-nematocidal B. thuringiensis strainDSM350, originally obtained from the German Collectionof Microorganisms and Cell Cultures (Deutsche Sammlungvon Mikroorganismen und Zellkulturen GmbH, DSMZ,Braunschweig, Germany); (iii) the pathogenic P. aeruginosastrain PA14, obtained from Dennis H. Kim, Boston, USA;and (iv) the non-pathogenic E. coli OP50. Before the startof this study, the three B. thuringiensis strains were culturedin large quantities as previously described [32]. The culturesconsisted mainly of spores associated to nematocidal toxinsin the case of B-18679 and B-18247, and non nematocidaltoxins in the case of the control DSM 350. All cultureswere set to a concentration of 1.5 × 1010 particles/ml,

Nakad et al. BMC Genomics (2016) 17:280 Page 14 of 20

assessed through standard Thoma counting chambers andmicroscopic analysis. The cultures were cryo-preserved inaliquots at –20 °C; spore viability and pathogenicity are notaffected by freezing under these conditions [38, 40, 72].Usage of viable spore aliquots from the same starting cul-ture allowed us to minimize variation across experiments,thus enhancing comparability of the data from differentstudy approaches. In all cases, aliquots were thawed dir-ectly before each experiment. The bacterial cultures werethen diluted, as indicated below in the description of thevarious assays.

Lawn leaving assaysThe assay was designed for 9 cm petri dishes with eitherPeptone free NGM (PFM) for B. thuringiensis assays orpeptone-rich NGM for PA14 assays. 30 μl of the testbacterium were spotted onto the center of the plate and80 μl of E. coli OP50 only were additionally placed inthe shape of a ring (Additional file 1: Figure S1). Thisring of OP50 protects escaping worms from starvation,minimizing their return to the lawn in the center. Thetest bacterium consisted of either B. thuringiensis dilutedwith OP50 at 1:250 from a stock with a concentration of1.5 × 1010 particles/ml, or PA14 diluted with OP50 in a4:1 ratio. 10 hermaphroditic fourth instar larvae (L4)were picked onto the test lawn. Experiments were per-formed at 20 °C.Leaving behavior was recorded by counting the num-

ber of worms on the lawn at different time points of ex-posure and calculated according to the followingformula:

Leaving index ¼ 10−number of worms on the lawn10

The screens of RILs (200 lines) and ILs (90 lines) weredone using a randomized block design on 17 dates, al-ways including the parental strains of these lines, N2and CB4856, as internal controls. Each C. elegans strain-bacteria-time point combination was assayed in threereplicates, resulting in a total of 35040 individual datapoints. The screens of npr-1 mutants included 12 repli-cates of each treatment.We would like to note that the leaving assay for PA14

was performed at 20 °C and thus at a different temperaturethan the standard PA14 survival assays at 25 °C (see belowand [19]). The reason is that the 25 °C temperature led toincreased bacterial growth on the assay plate, which causedenhanced dispersal of bacterial colonies through the crawl-ing worms, thus compromising reliable scoring of the leav-ing behavior. Such a bias was not observed at 20 °C. As ourresults did confirm previously published data on C. elegansavoidance behavior towards PA14 [19], our assay conditions

allowed us to characterize a robust behavioral responseagainst this pathogen.

Survival assaysFor survival analysis with B. thuringiensis, 6 cm peptonefree NGM plates were inoculated with 100 μl of a mix-ture of B. thuringiensis with E. coli OP50. Mixtures wereprepared in seven dilutions: 1:2, 1:5, 1:10, 1:30, 1:50,1:100, and 1:250 (equivalent to the relative concentrationgiven in the main text). Plates were left to dry overnight(9–15 h) at 20 °C. 30 L4 hermaphrodites were pickedonto each assay plates. After 24 h, survival was recordedby counting alive, dead and missing worms. Each treat-ment group was replicated 8 times across 8 runs (onereplicate per run).Analysis of P. aeruginosa effects was based on 3 cm

peptone-rich NGM plates, which were inoculated with5 μl of an overnight culture at 37 °C of either PA14 orOP50. Seeded plates were incubated overnight at 37 °Cand then at 25 °C. 30 L4 hermaphrodites were pickedonto each assay plate and stored at 25 °C in the dark.Alive and dead worms were scored every 24 h and sur-viving worms were transferred to new assay plates every48 h. Each treatment group was replicated 10 timesacross two runs.

Statistical analysis of phenotypic dataWe used the non-parametric Kruskall-Wallis test to as-sess differences in leaving behavior between the C. ele-gans strains, and a Bonferroni based adjustment tocorrect for multiple testing. We used Kaplan Meier ana-lysis applying the Log Rank test to assess differences inC. elegans’ survival on PA14. A Bonferroni based adjust-ment was used to correct for multiple testing. We usedGLM ordinal logistic regression analysis to assess differ-ences in survival between the C. elegans strains acrossthe concentration range of B. thuringiensis, using C. ele-gans strains, BT concentration and the interaction be-tween the two as factors. A Bonferroni based adjustmentwas used to correct for multiple testing. All escape andsurvival assays data were analyzed using the programJMP version 9.0 (SAS Institute Inc.), while graphic illus-trations were produced with the program SIGMAPLOTversion 12.0 (Systat Software Inc.).

Quantitative trait locus (QTL) analysisThe QTL analysis was performed on the average of threereplicates per genotype/line of the calculated proportion“leave” (see assay method) of 200 Recombinant InbredLines (RILs) [41, 55–58, 73–76]. QTLs were calculated bysingle marker mapping using a linear model (trait ~marker + error) for each marker using a custom writtenscript in the statistical programming language “R” [77].Significance levels were estimated from 1000 permutations

Nakad et al. BMC Genomics (2016) 17:280 Page 15 of 20

of the data. The analysis calculated the significance of thelinkage between the genetic marker and the trait one byone for each bacterium and exposure time point separ-ately. We furthermore evaluated epistatic interactionsbetween each two markers across the genome for eachbacterium separately using the following model: pheno-type ~ time +marker1 * marker2. BIN mapping in the setof 90 ILs was done as described in [42, 56].

RNA isolation and sequencingN2 and npr-1(ur89) worms were exposed to either PA14,BT B-18247 or OP50 for either 12 or 24 h of exposure. Theexperiment had 3 replicates of each treatment combination(a total of 36 samples across treatment combinations andreplicates). Exposure experiments were performed on largeAgar plates (15 cm diameter), which were fully coveredwith a bacterial lawn, thus minimizing escape responsesand resulting in comparable occupancy rates across thetreatment combinations. At both time points (12 & 24),worms were washed off the exposure plates with PBS con-taining 0.3 % Tween20® and resuspended in TRIzol® (LifeTechnologies) reagent. Prior to RNA extraction, worm sus-pensions were treated five times with a freeze-thaw cycleusing liquid nitrogen and a thermo block at 45 °C. RNAextraction was performed using a NucleoSpin® mRNA ex-traction kit (Macherey-Nagel). RNA samples were treatedwith DNAse, and then stored at –80 °C. RNA libraries wereprepared for sequencing using standard Illumina protocols.Libraries were sequenced on an Illumina HiSeq™ 2000sequencing machine with a paired-end strategy and readlength of 100 nucleotides.

Statistical analysis of transcriptomic dataRNAseq reads were mapped to the C. elegans genomefrom Wormbase version WS235 (www.wormbase.org)by Tophat2 [78] using option –b2-very-sensitive, otherdefault settings and without a transcriptome reference.Tophat2 aligns RNAseq reads to a genome based on theultra-fast short read mapping program Bowtie [79]. Esti-mation of transcript abundance and significantly differ-entially expressed genes were identified by Cuffdiff [80]using the quartile normalization method [81]. Cuffdiff isa program from the Cufflinks package and aims to findsignificant changes in transcript expression in consider-ation of possible formation of isoforms for a particulargene. The raw data is available from the GEO database[82, 83] under the GSE number GSE60063.For clustering and visualization, transcripts with a sig-

nificant change between different conditions (adjusted p-value < 0.01 by the false discovery rate, FDR) were treatedas signature for each comparison. Due to the biologicalvariation of the replicates, the p-value, instead of fold-change, of those genes were firstly log10-transformed andordered according to increasing or decreasing expression

and then taken as input for k-means cluster analysis usingcluster 3.0 [84] with a k of 8. A heatmap was generated byTreeView version 1.1.4r3 [85]. Principal component ana-lysis (PCA) was carried out on log-transformed gene ex-pression profiles using a probabilistic PCA algorithm [86]from R package pcaMethods [87], which links PCA to theprobability density of patterns. Dimensionality of the sam-ples was reduced from 57165 (total isoforms) to threedimensions (PCs). Motif analysis was carried out on thepromoter regions, -600 bp and 250 bp relative to tran-scription start sites (TSS), of genes in each group. De novomotif discovery was performed by AMD [54].

GO and EASE analysisGene ontology (GO) and a gene set enrichment analysiswas carried out on each group of genes from the K meanscluster analysis. GO analysis was performed using DAVID[52, 53] with a cut-off of p-value < 0.05, adjusted by FDR.For the gene set analysis, we used EASE [50], a free,stand-alone software package from DAVID bioinformaticsresources (http://david.abcc.ncifcrf.gov/). As recently de-scribed [51, 64], we constructed a C. elegans-specific geneset database, WormExp [51], from published data and alsousing the previously established data sets collected by IlkaEngelmann et al., [88]. Based on this data set, we per-formed the EASE analysis and selected the results with aBonferroni adjusted p-value < 0.05.

eQTL enrichment analysiseQTL enrichment was done using a hypergeometric testin R. The eQTL sets [41, 57–59] were obtained fromWormQTL.org [55, 60]. The eQTLs at a specific locuswere compared to the genes in a specific expressioncluster, as identified from the above described K-meanscluster analysis.

Ethics approval and consent to participateNot applicable.

Consent for publicationNot applicable.

Availability of data and materialThe datasets, supporting the conclusions of this article,is available in case of the QTL analysis from WormQTL[55, 60], in case of the phenotypic analysis of N2, CB4856,and the npr-1 mutants in Additional file 2 of this article,and in case of the transcriptome analysis from the GEOdatabase [82, 83] under the GSE number GSE60063. TheC. elegans strains are available from the CaenorhabditisGenetic Center (CGC), which is funded by NIH Office ofResearch Infrastructure Programs (P40 OD010440). Allbacterial strains are available from the correspondingauthor upon request.

Nakad et al. BMC Genomics (2016) 17:280 Page 16 of 20

Additional files

Additional file 1: Illustration of the lawn leaving assay. 10 hermaphroditesat the L4 stage were transferred by picking onto 9 cm peptone free NGMplates containing a lawn of the tested bacteria, each mixed with E. coliOP50 and surrounded by a ring of 80 μl of OP50. (TIF 144 kb)

Additional file 2: Raw data for the analysis of N2, CB4856, and the npr-1 mutants. Sheet 1, Lawn leaving behavior of N2 and CB4856 from Bacillusthuringiensis - Raw data for Fig. 1; sheet 2, Lawn leaving behavior ofmutants from Bacillus thuringiensis - Raw data for Fig. 4; sheet 3, Surivalof N2 and CB4856 on Bacillus thuringiensis - Raw data for Additionalfile 9; sheet 4, Surival of mutants on Bacillus thuringiensis - Raw data forFig. 5; sheet 5, Lawn leaving behavior of N2 and CB4856 fromPseudomonas aeruginosa PA14 - Raw data for Additional file 10; sheet 6,Lawn leaving behavior of mutants from Pseudomonas aeruginosa PA14 -Raw data for Fig. 6; sheet 7, Surival of mutants on Pseudomonas PA14 - Rawdata for Fig. 5. (XLS 482 kb)

Additional file 3: Table on the statistical results for the comparisonbetween N2 and CB4856 leaving behavior towards B. thuringiensis and E. coli.(PDF 83 kb)

Additional file 4: Table on the statistical results for the pairwisecomparisons of the leaving response on the E.coli OP50 control versusthe B. thuringiensis strains. (PDF 87 kb)

Additional file 5: Figure on the leaving phenotypes of the introgressionlines (ILs) plotted against the introgression position along the chromosomes.(A) Results for E. coli strain OP50; (B) non-nematocidal B. thuringiensis strainDSM350; (C) nematocidal B. thuringiensis B-18247; and (D) highly nematocidalB. thuringiensis B-18679. Green and red lines show the results after either14 h or 24 h exposure, respectively. Position of markers is given alongthe X axis. Light gray vertical lines indicate boundaries of the chromosomes.(TIF 1279 kb)

Additional file 6: Table on the statistical results for the comparison ofthe N2 and CB4856 leaving behavior with that of the mutant strainstowards B. thuringiensis and E. coli. (PDF 96 kb)

Additional file 7: Table on the statistical results for the separatecomparison between C. elegans N2 and CB4856 survival on nematocidalB. thuringiensis. (PDF 74 kb)

Additional file 8: Table on the statistical results for the pairwise comparisonof survival of the C. elegans strains on nematocidal B. thuringiensis. (PDF 78 kb)

Additional file 9: Figure on the separate analysis of N2 and CB4856survival in the presence of nematocidal B. thuringiensis. (A) Survival on B.thuringiensis strain B-18247; and (B) B-18679. Survival on the Y axis is plottedagainst BT concentration on the X axis. Error bars represent standard errorof the means. The statistical results are given in Additional file 6. (TIF 106 kb)

Additional file 10: Figure on the separate analysis of lawn leavingbehavior of N2 and CB4856 towards E. coli and P. aeruginosa. (A) Resultsfor avoidance of E. coli strain OP50; and (B) P. aeruginosa strain PA14. Theasterisk (*) points to a significant difference to N2. The dotted referenceline indicates the 0.5 avoidance response. Statistical results are shown inAdditional file 10. (TIF 97 kb)

Additional file 11: Table on the statistical results for the comparisonbetween N2 and CB4856 leaving behavior towards E. coli and P.aeruginosa. (PDF 76 kb)

Additional file 12: Table on the statistical results for the comparison ofthe N2 and CB4856 leaving behavior with that of the mutant strainstowards P. aeruginosa and E. coli. (PDF 90 kb)

Additional file 13: Table on the statistical results for the comparison ofC. elegans survival rate on P. aeruginosa PA14 versus the control E. coliOP50. (PDF 74 kb)

Additional file 14: Table on the statistical results for the pairwisecomparisons of C. elegans survival rates on P. aeruginosa PA14. (PDF 72 kb)

Additional file 15: Table with the list of differentially expressed genes

after exposure of the C. elegans N2 wildtype or npr-1 mutant to

pathogenic B. thuringiensis B-18247, pathogenic P. aeruginosa PA14, or

non-pathogenic E. coli. (XLS 2096 kb)

Additional file 16: Results of the eQTL enrichment analysis of the Kmeans clusters of differentially expressed genes. (XLS 160 kb)

Additional file 17: Results of the GO term enrichment analysis (firstsheet), the motif analysis (second sheet), and the C. elegans-specific EASEenrichment analysis (third sheet) of the K means clusters of differentiallyexpressed genes. (XLS 316 kb)

AbbreviationsAMD: automated motif discovery; DAVID: database for annotation, visualizationand integrated discovery; EASE: expression analysis systematic explorer;eQTL: expression quantitative trait locus; FDR: false discovery rate; GEO: geneexpression omnibus; GLM: generalized linear model; GO: gene ontology;PCA: principal component analysis; QTL: quantitative trait locus.

Competing interestsThe authors declare that they have no competing interests.

Authors’ contributionsRN conceived the study, performed and participated in all experiments,analyzed the data, and drafted the manuscript. LBS conceived and performedQTL analysis and drafted the manuscript. WY performed transcriptomic analysisand drafted the manuscript. SE, FS, TGM, LR, ACM, KD helped performingphenotypic and functional genetic experiments. PCR conceived andparticipated in the transcriptomic analysis. JEK provided the RIL and ILlibraries, conceived the QTL analysis, and drafted the manuscript. HS conceivedthe study, analyzed the data, and drafted the manuscript. All authors read andapproved the final manuscript.

AcknowledgementsWe are grateful to all members of the Schulenburg and Kammenga labs forsupport and advice. We thank particularly Christiana Anagnostou, DanielaHaase, Leila Masri, Barbara Pees, Joost Riksen, Stefanie Rohwer, and AnnaSheppard. We also thank the Kiel ICMB sequencing team (especially MarkusSchilhabel, Melanie Friskovec, Melanie Schlapkohl) for technical assistance.The C. elegans strains were provided by the Caenorhabditis Genetic Center(CGC), which is funded by NIH Office of Research Infrastructure Programs(P40 OD010440). We thank the CGC for providing these strains and Dennis Kimfor providing the P. aeruginosa strain PA14.

FundingWe are most grateful for financial support from the German Science Foundationto HS (DFG grants SCHU 1415/6 and SCHU 1415/9); Kiel University to HS and KD;infra-structural funds from the DFG Excellence Cluster Inflammation at Interfacesto HS and PR; the Netherlands Organisation for Scientific Research to LBS and JEK(grants 823.01.001 and 855.01.151); the HFSP to JEK; and the International MaxPlanck Research School (IMPRS) for Evolutionary Biology to WY.

Author details1Department of Evolutionary Ecology and Genetics, Zoological Institute,University of Kiel, 24098 Kiel, Germany. 2Cologne Excellence Cluster forCellular Stress Responses in Ageing-Associated Diseases (CECAD) andSystems Biology of Ageing, University of Cologne, Joseph-Stelzmann-Str. 26,50931 Cologne, Germany. 3Laboratory of Nematology, WageningenUniversity, Wageningen 6708 PB, The Netherlands. 4Institute for ClinicalMolecular Biology, University of Kiel, 24098 Kiel, Germany.

Received: 7 November 2015 Accepted: 25 March 2016

References1. Schulenburg H, Boehnisch C, Michiels NK. How do invertebrates

generate a highly specific innate immune response? Mol Immunol. 2007;44(13):3338–44.

2. Irazoqui JE, Urbach JM, Ausubel FM. Evolution of host innate defence:insights from Caenorhabditis elegans and primitive invertebrates. Nat RevImmunol. 2010;10(1):47–58.

3. Buchon N, Silverman N, Cherry S. Immunity in Drosophila melanogaster -from microbial recognition to whole-organism physiology. Nat RevImmunol. 2014;14(12):796–810.

Nakad et al. BMC Genomics (2016) 17:280 Page 17 of 20

4. Meisel JD, Kim DH. Behavioral avoidance of pathogenic bacteria byCaenorhabditis elegans. Trends Immunol. 2014;35(10):465–70.

5. Ewbank JJ, Pujol N. Local and long-range activation of innate immunity byinfection and damage in C. elegans. Curr Opin Immunol. 2016;38:1–7.

6. Luhachack LG, Visvikis O, Wollenberg AC, Lacy-Hulbert A, Stuart LM, IrazoquiJE. EGL-9 controls C. elegans host defense specificity through prolylhydroxylation-dependent and-independent HIF-1 pathways. PLoS Pathog.2012;8(7):e1002798.

7. Shao Z, Zhang Y, Ye Q, Saldanha JN, Powell-Coffman JA. C. elegans SWAN-1Binds to EGL-9 and regulates HIF-1-mediated resistance to the bacterialpathogen Pseudomonas aeruginosa PAO1. PLoS Pathog. 2010;6(8):e1001075.

8. Bellier A, Chen C-S, Kao C-Y, Cinar HN, Aroian RV. Hypoxia and the hypoxicresponse pathway protect against pore-forming toxins in C. elegans.PLoS Pathog. 2009;5(12):e1000689.

9. Tenor JL, Aballay A. A conserved Toll-like receptor is required forCaenorhabditis elegans innate immunity. EMBO Rep. 2008;9(1):103–9.

10. Pujol N, Link EM, Liu LX, Kurz CL, Alloing G, Tan M-W, Ray KP, Solari R,Johnson CD, Ewbank JJ. A reverse genetic analysis of components ofthe Toll signaling pathway in Caenorhabditis elegans. Curr Biol. 2001;11(11):809–21.

11. Schulenburg H, Ewbank JJ. The genetics of pathogen avoidance inCaenorhabditis elegans. Mol Microbiol. 2007;66(3):563–70.

12. Petersen C, Dirksen P, Schulenburg H. Why we need more ecology forgenetic models such as C. elegans. Trends Genet. 2015;31:120–127.

13. Felix MA, Braendle C. The natural history of Caenorhabditis elegans. Curr Biol.2010;20(22):R965–9.

14. Felix M-A, Duveau F. Population dynamics and habitat sharing of naturalpopulations of Caenorhabditis elegans and C. briggsae. BMC Biol. 2012;10:59.

15. Petersen C, Dirksen P, Prahl S, Strathmann EA, Schulenburg H. Theprevalence of Caenorhabditis elegans across 1.5 years in selected NorthGerman locations: the importance of substrate type, abiotic parameters, andCaenorhabditis competitors. BMC Ecol. 2014;14:4.

16. Pradel E, Zhang Y, Pujol N, Matsuyama T, Bargmann CI, Ewbank JJ.Detection and avoidance of a natural product from the pathogenicbacterium Serratia marcescens by Caenorhabditis elegans. Proc Natl Acad SciU S A. 2007;104(7):2295–300.

17. Reddy KC, Andersen EC, Kruglyak L, Kim DH. A polymorphism in npr-1 is abehavioral determinant of pathogen susceptibility in C. elegans. Science.2009;323(5912):382–4.

18. Styer KL, Singh V, Macosko E, Steele SE, Bargmann CI, Aballay A. Innateimmunity in Caenorhabditis elegans is regulated by neurons expressing NPR-1/GPCR. Science. 2008;322(5900):460–4.

19. Chang HC, Paek J, Kim DH. Natural polymorphisms in C. elegans HECW-1 E3ligase affect pathogen avoidance behaviour. Nature. 2012;480(7378):525–9.

20. McMullan R, Anderson A, Nurrish S. Behavioral and immune responses toinfection require Gαq- RhoA Signaling in C. elegans. PLoS Pathog. 2012;8(2):e1002530.

21. Hasshoff M, Boehnisch C, Tonn D, Hasert B, Schulenburg H. The role ofCaenorhabditis elegans insulin-like signaling in the behavioral avoidance ofpathogenic Bacillus thuringiensis. FASEB J. 2007;21(8):1801–12.

22. Zhang Y, Lu H, Bargmann CI. Pathogenic bacteria induce aversiveolfactory learning in Caenorhabditis elegans. Nature. 2005;438(7065):179–84.

23. Schulenburg H, Muller S. Natural variation in the response of Caenorhabditiselegans toward Bacillus thruingiensis. Parasitology. 2004;128:433–43.

24. Volkers RJM, Snoek LB, van Hellenberg Hubar CJ, Coopman R, Chen W,Yang W, Sterken MG, Schulenburg H, Braeckman BP, Kammenga JE. Gene-environment and protein-degradation signatures characterize genomic andphenotypic diversity in wild Caenorhabditis elegans populations. BMC Biol.2013;11(1):93.

25. Sicard M, Hering S, Schulte R, Gaudriault S, Schulenburg H. The effect ofPhotorhabdus luminescens (Enterobacteriaceae) on the survival,development, reproduction and behaviour of Caenorhabditis elegans(Nematoda: Rhabditidae). Environ Microbiol. 2007;9(1):12–25.

26. Glater EE, Rockman MV, Bargmann CI. Multigenic natural variation underliesCaenorhabditis elegans olfactory preference for the bacterial pathogenSerratia marcescens. G3. 2014;4(2):265–76.

27. Andersen EC, Gerke JP, Shapiro JA, Crissman JR, Ghosh R, Bloom JS, Felix M-A,Kruglyak L. Chromosome-scale selective sweeps shape Caenorhabditis elegansgenomic diversity. Nat Genet. 2012;44(3):285–90.

28. de Bono M, Bargmann CI. Natural variation in a neuropeptide Y receptorhomolog modifies social behavior and food response in C. elegans. Cell.1998;94(5):679–89.

29. Rogers C, Persson A, Cheung B, de Bono M. Behavioral motifs and neuralpathways coordinating O2 responses and aggregation in C. elegans.Curr Biol. 2006;16(7):649–59.

30. Weber KP, De S, Kozarewa I, Turner DJ, Babu MM, de Bono M. Wholegenome sequencing highlights genetic changes associated with laboratorydomestication of C. elegans. PLoS One. 2010;5(11):e13922.

31. Bendesky A, Tsunozaki M, Rockman MV, Kruglyak L, Bargmann CI.Catecholamine receptor polymorphisms affect decision-making in C.elegans. Nature. 2012;472:313–318.

32. Borgonie G, Van Driessche R, Leyns F, Arnaut G, De Waele D, Coomans A.Germination of Bacillus thuringiensis spores in bacteriophagous nematodes(Nematoda: Rhabditida). J Invertebr Pathol. 1995;65(1):61–7.

33. Borgonie G, Claeys M, Leyns F, Arnaut G, De Waele D, Coomans AV. Effectof a nematicidal Bacillus thuringiensis strain on free-living nematodes. 3.Characterization of the intoxication process. Fundam Appl Nematol. 1996;19:523–8.

34. Wei JZ, Hale K, Carta L, Platzer E, Wong C, Fang SC, Aroian RV. Bacillusthuringiensis crystal proteins that target nematodes. Proc Natl Acad SciU S A. 2003;100(5):2760–5.

35. Griffitts JS, Aroian RV. Many roads to resistance: how invertebrates adapt toBt toxins. Bioessays. 2005;27(6):614–24.

36. Nielsen-LeRoux C, Gaudriault S, Ramarao N, Lereclus D, Givaudan A. Howthe insect pathogen bacteria Bacillus thuringiensis and Xenorhabdus/Photorhabdus occupy their hosts. Curr Opin Microbiol. 2012;15(3):220–31.

37. Wang J, Nakad R, Schulenburg H. Activation of the Caenorhabditis elegansFOXO family transcription factor DAF-16 by pathogenic Bacillus thuringiensis.Dev Comp Immunol. 2012;37(1):193–201.

38. Schulte RD, Makus C, Hasert B, Michiels NK, Schulenburg H. Host–parasitelocal adaptation after experimental coevolution of Caenorhabditis elegansand its microparasite Bacillus thuringiensis. Proc Royal Soc B. 2011;278(1719):2832–9.

39. Schulte RD, Makus C, Hasert B, Michiels NK, Schulenburg H. Multiplereciprocal adaptations and rapid genetic change upon experimentalcoevolution of an animal host and its microbial parasite. Proc NatlAcad Sci. 2010;107(16):7359–64.

40. Masri L, Branca A, Sheppard AE, Papkou A, Laehnemann D, Guenther PS,Prahl S, Saebelfeld M, Hollensteiner J, Liesegang H. Host-pathogencoevolution: the selective advantage of Bacillus thuringiensis virulence andits cry toxin genes. PLoS Biol. 2015;13(6):e1002169.

41. Li Y, Ãlvarez OA, Gutteling EW, Tijsterman M, Fu J, Riksen JAG, Hazendonk E,Prins P, Plasterk RHA, Jansen RC, et al. Mapping determinants of gene expressionplasticity by genetical genomics in C. elegans. PLoS Genet. 2006;2(12):e222.

42. Doroszuk A, Snoek LB, Fradin E, Riksen J, Kammenga J. A genome-widelibrary of CB4856/N2 introgression lines of Caenorhabditis elegans. NucleicAcids Res. 2009;37(16):e110.

43. Balla KM, Troemel ER. Caenorhabditis elegans as a model for intracellularpathogen infection. Cell Microbiol. 2015;15(8):1313–22.

44. Wilfert L, Gadau J, Schmid‐Hempel P. The genetic architecture of immunedefense and reproduction in male Bombus terrestris bumblebees. Evolution.2007;61(4):804–15.

45. Luijckx P, Fienberg H, Duneau D, Ebert D. A matching-allele model explainshost resistance to parasites. Curr Biol. 2013;23(12):1085–8.

46. Schulenburg H, Kurtz J, Moret Y, Siva-Jothy MT. Introduction. Ecologicalimmunology. Philo Transac Royal Soc B. 2009;364(1513):3–14.

47. Ferrandon D, Imler J-L, Hetru C, Hoffmann JA. The Drosophila systemicimmune response: sensing and signalling during bacterial and fungalinfections. Nat Rev Immunol. 2007;7(11):862–74.

48. Cheung BHH, Cohen M, Rogers C, Albayram O, de Bono M. Experience-dependent modulation of C. elegans behavior by ambient oxygen. Curr Biol.2005;15(10):905–17.

49. Chang AJ, Chronis N, Karow DS, Marletta MA, Bargmann CI. A distributedchemosensory circuit for oxygen preference in C. elegans. PLoS Biol. 2006;4(9):e274.

50. Hosack DA, Dennis Jr G, Sherman BT, Lane HC, Lempicki RA. Identifyingbiological themes within lists of genes with EASE. Genome Biol. 2003;4(10):R70.

51. Yang W, Dierking K, Schulenburg H. WormExp: a web-based application fora Caenorhabditis elegans-specific gene expression enrichment analysis.Bioinformatics. 2015; [Epub ahead of print].

Nakad et al. BMC Genomics (2016) 17:280 Page 18 of 20

52. Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysisof large gene lists using DAVID bioinformatics resources. Nat Protoc. 2008;4(1):44–57.

53. Huang DW, Sherman BT, Lempicki RA. Bioinformatics enrichment tools:paths toward the comprehensive functional analysis of large gene lists.Nucleic Acids Res. 2009;37(1):1–13.

54. Shi J, Yang W, Chen M, Du Y, Zhang J, Wang K. AMD, an automated motifdiscovery tool using stepwise refinement of gapped consensuses. PLoSOne. 2011;6(9):e24576.

55. Snoek LB, Van der Velde KJ, Arends D, Li Y, Beyer A, Elvin M, Fisher J, HajnalA, Hengartner MO, Poulin GB, et al. Worm QTL—public archive and analysisweb portal for natural variation data in Caenorhabditis spp. Nucleic AcidsRes. 2013;41(D1):D738–43.

56. Snoek LB, Orbidans HE, Stastna JJ, Aartse A, Rodriguez M, Riksen JAG,Kammenga JE, Harvey SC. Widespread genomic incompatibilities inCaenorhabditis elegans. G3. 2014;4(10):1813–23.

57. Li Y, Breitling R, Snoek LB, Velde KJ, Swertz MA, Riksen JAG, Jansen RC,Kammenga JE. Global genetic robustness of the alternative splicingmachinery in Caenorhabditis elegans. Genetics. 2010;186:405–10.

58. Viñuela A, Snoek LB, Riksen JAG, Kammenga JE. Genome-wide geneexpression regulation as a function of genotype and age in C. elegans.Genome Res. 2010;20(7):929–37.

59. Rockman MV, Skrovanek SS, Kruglyak L. Selection at linked sitesshapes heritable phenotypic variation in C. elegans. Science. 2010;330(6002):372–6.

60. Snoek LB, Joeri van der Velde K, Li Y, Jansen RC, Swertz MA, Kammenga JE.Worm variation made accessible: Take your shopping cart to store, link, andinvestigate! In: Worm. Abingdon, UK: Taylor & Francis; 2014: e28357.

61. Miltsch SM, Seeberger PH, Lepenies B. The C-type lectin-like domaincontaining proteins CLEC-39 and CLEC-49 are crucial for Caenorhabditiselegans immunity against Serratia marcescens infection. Dev ComparImmunol. 2014;45(1):67–73.

62. Irazoqui JE, Troemel ER, Feinbaum RL, Luhachack LG, Cezairliyan BO,Ausubel FM. Distinct pathogenesis and host responses during infection ofC. elegans by P. aeruginosa and S. aureus. PLoS Pathog. 2010;6(7):e1000982.

63. O’Rourke D, Baban D, Demidova M, Mott R, Hodgkin J. Genomic clusters,putative pathogen recognition molecules, and antimicrobial genes areinduced by infection of C. elegans with M. nematophilum. Genome Res.2006;16(8):1005–16.

64. Yang W, Dierking K, Esser D, Tholey A, Leippe M, Rosenstiel P, Schulenburg H.Overlapping and unique signatures in the proteomic and transcriptomicresponses of the nematode Caenorhabditis elegans toward pathogenic Bacillusthuringiensis. Dev Compar Immunol. 2015;51(1):1–9.

65. Kim DH, Feinbaum R, Alloing G, Emerson FE, Garsin DA, Inoue H, Tanaka-Hino M,Hisamoto N, Matsumoto K, Tan M-W, et al. A conserved p38 MAP kinasepathway in Caenorhabditis elegans innate immunity. Science. 2002;297(5581):623–6.

66. Shapira M, Hamlin BJ, Rong J, Chen K, Ronen M, Tan M-W. A conserved rolefor a GATA transcription factor in regulating epithelial innate immuneresponses. Proc Natl Acad Sci. 2006;103(38):14086–91.

67. Kao C-Y, Los FCO, Huffman DL, Wachi S, Kloft N, Husmann M, Karabrahimi V,Schwartz J-L, Bellier A, Ha C, et al. Global functional analyses of cellularresponses to pore-forming toxins. PLoS Pathog. 2011;7(3):e1001314.

68. Feng N, Zhi D, Zhang L, Tian J, Ren H, Li C, Zhu H, Li H. Molecularmechanisms of resistance to human pathogenic bacteria in Caenorhabditiselegans by MEV-1 mediated oxidative stress. Biochem Biophys Res Commun.2015;459(3):481–7.

69. Schmid-Hempel P. Parasite immune evasion: a momentous molecular war.Trends Ecol Evol. 2008;23(6):318–26.

70. Niu Q, Huang X, Zhang L, Xu J, Yang D, Wei K, Niu X, An Z, Bennett JW, Zou C.A Trojan horse mechanism of bacterial pathogenesis against nematodes. ProcNatl Acad Sci. 2010;107(38):16631–6.

71. Stiernagle T. Maintenance of C. elegans. WormBook. The C. elegans researchcommunity. In: WormBook. 2006.

72. Leyns F, Borgonie G, Arnaut G, De Waele D. Nematicidal activity of Bacillusthuringiensis isolates. Fundam Appl Nematol. 1995;18(3):211–8.

73. Gutteling EW, Doroszuk A, Riksen JAG, Prokop Z, Reszka J, Kammenga JE.Environmental influence on the genetic correlations between life-historytraits in Caenorhabditis elegans. Heredity. 2007;98(4):206–13.

74. Viñuela A, Snoek LB, Riksen JAG, Kammenga JE. Aging uncouples heritabilityand expression-QTL in Caenorhabditis elegans. G3. 2012;2(5):597–605.

75. Elvin M, Snoek L, Frejno M, Klemstein U, Kammenga J, Poulin G. A fitnessassay for comparing RNAi effects across multiple C. elegans genotypes. BMCGenomics. 2011;12(1):1–14.

76. Rodriguez M, Snoek LB, Riksen JAG, Bevers RP, Kammenga JE. Geneticvariation for stress-response hormesis in C. elegans lifespan. Exp Gerontol.2012;47(8):581–7.

77. R: A Language and Environment for Statistical Computing, R Core Team, RFoundation for Statistical Computing, Vienna, Austria, 2015, http://www.R-project.org/. 2015.

78. Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg SL. TopHat2:accurate alignment of transcriptomes in the presence of insertions,deletions and gene fusions. Genome Biol. 2013;14(4):R36.

79. Langmead B, Trapnell C, Pop M, Salzberg SL. Ultrafast and memory-efficientalignment of short DNA sequences to the human genome. Genome Biol.2009;10(3):R25.

80. Trapnell C, Hendrickson DG, Sauvageau M, Goff L, Rinn JL, Pachter L.Differential analysis of gene regulation at transcript resolution with RNA-seq.Nat Biotechnol. 2013;31(1):46–53.

81. Robinson MD, Oshlack A. A scaling normalization method for differentialexpression analysis of RNA-seq data. Genome Biol. 2010;11(3):R25.

82. Edgar R, Domrachev M, Lash AE. Gene expression omnibus: NCBI geneexpression and hybridization array data repository. Nucleic Acids Res. 2002;30(1):207–10.

83. Barrett T, Wilhite SE, Ledoux P, Evangelista C, Kim IF, Tomashevsky M, Marshall KA,Phillippy KH, Sherman PM, Holko M. NCBI GEO: archive for functional genomicsdata sets update. Nucleic Acids Res. 2013;41(D1):D991–5.

84. de Hoon MJ, Imoto S, Nolan J, Miyano S. Open source clustering software.Bioinformatics. 2004;20(9):1453–4.

85. Saldanha AJ. Java Treeview–extensible visualization of microarray data.Bioinformatics. 2004;20(17):3246–8.

86. Tipping ME, Bishop CM. Probabilistic principal component analysis. J RoyalStat Soc Ser B. 1999;61(3):611–22.

87. Stacklies W, Redestig H, Scholz M, Walther D, Selbig J. pcaMethods–abioconductor package providing PCA methods for incomplete data.Bioinformatics. 2007;23(9):1164–7.

88. Engelmann I, Griffon A, Tichit L, Montanana-Sanchis F, Wang G, Reinke V,Waterston RH, Hillier LW, Ewbank JJ. A comprehensive analysis of geneexpression changes provoked by bacterial and fungal infection in C.elegans. PLoS One. 2011;6(5):e19055.

89. Huffman DL, Abrami L, Sasik R, Corbeil J, van der Goot FG, Aroian RV.Mitogen-activated protein kinase pathways defend against bacterial pore-forming toxins. Proc Natl Acad Sci U S A. 2004;101(30):10995–1000.

90. Boehnisch C, Wong D, Habig M, Isermann K, Michiels NK, Roeder T, May RC,Schulenburg H. Protist-type lysozymes of the nematode Caenorhabditiselegans contribute to resistance against pathogenic Bacillus thuringiensis.PLoS One. 2011;6(9):e24619.

91. Troemel ER, Chu SW, Reinke V, Lee SS, Ausubel FM, Kim DH. p38 MAPKregulates expression of immune response genes and contributes tolongevity in C. elegans. PLoS Genet. 2006;2(11):e183.

92. Park SK, Tedesco PM, Johnson TE. Oxidative stress and longevity inCaenorhabditis elegans as mediated by SKN-1. Aging Cell. 2009;8(3):258–69.

93. Rohlfing A-K, Miteva Y, Hannenhalli S, Lamitina T. Genetic andphysiological activation of osmosensitive gene expression mimicstranscriptional signatures of pathogen infection in C. elegans. PLoS One.2010;5(2):e9010.

94. Cui Y, McBride SJ, Boyd WA, Alper S, Freedman JH. Toxicogenomic analysisof Caenorhabditis elegans reveals novel genes and pathways involved in theresistance to cadmium toxicity. Genome Biol. 2007;8(6):R122.

95. Murphy CT, McCarroll SA, Bargmann CI, Fraser A, Kamath RS, Ahringer J, Li H,Kenyon C. Genes that act downstream of DAF-16 to influence the lifespan ofCaenorhabditis elegans. Nature. 2003;424(6946):277–83.

96. McElwee JJ, Schuster E, Blanc E, Thomas JH, Gems D. Shared transcriptionalsignature in Caenorhabditis elegans Dauer larvae and long-lived daf-2mutants implicates detoxification system in longevity assurance. J BiolChem. 2004;279(43):44533–43.

97. Halaschek-Wiener J, Khattra JS, McKay S, Pouzyrev A, Stott JM, Yang GS, Holt RA,Jones SJM, Marra MA, Brooks-Wilson AR. Analysis of long-lived C. elegans daf-2mutants using serial analysis of gene expression. Genome Res. 2005;15(5):603–15.

98. Kim SK, Lund J, Kiraly M, Duke K, Jiang M, Stuart JM, Eizinger A, Wylie BN,Davidson GS. A gene expression map for Caenorhabditis elegans. Science.2001;293(5537):2087–92.

Nakad et al. BMC Genomics (2016) 17:280 Page 19 of 20

99. Menzel R, Bogaert T, Achazi R. A systematic gene expression screen ofCaenorhabditis elegans cytochrome P450 genes reveals CYP35 as stronglyxenobiotic inducible. Arch Biochem Biophys. 2001;395(2):158–68.

100. Reichert K, Menzel R. Expression profiling of five different xenobiotics usinga Caenorhabditis elegans whole genome microarray. Chemosphere. 2005;61(2):229–37.

101. Grove CA, De Masi F, Barrasa MI, Newburger DE, Alkema MJ, Bulyk ML,Walhout AJM. A multiparameter network reveals extensive divergencebetween C. elegans bHLH transcription factors. Cell. 2009;138(2):314–27.

102. McGhee JD, Sleumer MC, Bilenky M, Wong K, McKay SJ, Goszczynski B, Tian H,Krich ND, Khattra J, Holt RA. The ELT-2 GATA-factor and the global regulationof transcription in the C. elegans intestine. Dev Biol. 2007;302(2):627–45.

103. Niu W, Lu ZJ, Zhong M, Sarov M, Murray JI, Brdlik CM, Janette J, Chen C,Alves P, Preston E. Diverse transcription factor binding features revealed bygenome-wide ChIP-seq in C. elegans. Genome Res. 2011;21(2):245–54.

• We accept pre-submission inquiries

• Our selector tool helps you to find the most relevant journal

• We provide round the clock customer support

• Convenient online submission

• Thorough peer review

• Inclusion in PubMed and all major indexing services

• Maximum visibility for your research

Submit your manuscript atwww.biomedcentral.com/submit

Submit your next manuscript to BioMed Central and we will help you at every step:

Nakad et al. BMC Genomics (2016) 17:280 Page 20 of 20